E-Discovery Obligations Explained: Legal Holds, Privilege Review and Spoliation Sanctions

E-discovery is the process and the legal duty governing electronically stored information (ESI) in litigation: once litigation is reasonably anticipated, parties must preserve relevant ESI — email, messages, documents, and the systems that hold them — must collect and produce it in defensible ways under Federal Rule of Civil Procedure 34 and the proportionality limits of Rule 26, and must protect privileged material through a review process whose failures are waivable. The duty's teeth are the spoliation doctrine: destroying evidence after the preservation duty attaches draws sanctions up to an adverse-inference instruction — telling the jury it may assume the destroyed evidence was damaging — and, in bad-faith cases, dismissal or default judgment under Rule 37(e).

USA Post publishes information about litigation procedure, not legal advice.

When does the preservation duty attach, and what must be preserved?

When litigation is reasonably anticipated — a demand letter, an agency subpoena, a credible threat — which is routinely earlier than the complaint. The instrument is the legal hold: a written notice to custodians suspending routine deletion and identifying the subject matter; holds must be tracked, acknowledged, refreshed as facts develop, and — the recurring failure — extended to new custodians and to messaging platforms. Scope covers relevant ESI in the party's possession, custody or control, including employees' business communications on personal devices and messaging apps where the company's policies reach them. The 2022-2025 enforcement wave over ephemeral messaging — regulatory investigations and private litigation both — punished companies whose hold notices did not suspend auto-deleting apps, and courts have held that a hold which ignores the company's own WhatsApp/Signal usage is unreasonable as a matter of law in several antitrust and employment decisions.

What does a defensible collection and production look like?

Documented process at every step. Collection: targeted custodians, date ranges and keyword or technology-assisted searches agreed in the Rule 26(f) meet-and-confer, executed with forensics-grade tools that preserve metadata; self-collection by custodians is disfavored for exactly the credibility risk it creates. Processing and review: deduplication, threading, and a privilege review — attorney-client and work-product material withheld on a privilege log under Rule 26(b)(5), a task now split between human review and predictive coding/TAR, whose judicial acceptance (from Da Silva Moore forward) is settled. Production format: searchable metadata-bearing load files unless otherwise agreed. The proportionality balance of Rule 26(b)(1) limits burdensome requests — cost-shifting and sampling are the negotiation levers — but proportionality never excuses destroying what was preserved.

How do sanctions work under Rule 37(e)?

In two tiers. Curative measures: where ESI lost through failure to take reasonable steps cannot be restored or replaced, the court may order measures no greater than necessary to cure prejudice. Severe sanctions: only on prejudice or intent to deprive — the 2015 amendments to Rule 37(e) channeled spoliation doctrine, requiring a finding that a party acted with intent to deprive before adverse-inference instructions, dismissal or default. The practical battlegrounds: intent findings from circumstantial evidence (the employee who wiped a phone the week of the hold; the off-boarding process that deleted a departing executive's mailbox despite the hold), and the reasonableness of preservation steps measured against company size and ESI volume. State courts apply parallel doctrines; the Second Circuit's living standard from West v. Goodyear and CAT3 v. CableKinetics-line decisions continues to influence outcomes alongside the federal rule.

What about privilege failures — clawback and quick peek?

Producing privileged material is usually not fatal if the process was reasonable: Rule 26(b)(5)(B) lets a producing party claw back inadvertently produced privileged documents, and FRE 502(d) orders — entered in virtually every federal case at discovery's start — prevent waiver for productions made with reasonable precautions. What defeats protection: no review process at all, failure to act promptly on discovering production, and systemic carelessness. The high-stakes modern variant is privilege logging of AI-assisted review: courts have required disclosure of TAR methodologies in some disputes, and the 2024-2025 emergence of generative-AI review tools raises the same reasonableness questions early TAR faced — a documented, validated workflow is the defense.

How should companies prepare before litigation?

1. Map the ESI estate now — data sources, retention schedules, messaging platforms — because the first week of a hold is too late to discover systems.
2. Adopt retention policies with litigation-hold exceptions and actually enforce messaging-app governance; the gap between written policy and practice is Exhibit A in spoliation motions.
3. Pre-select e-discovery counsel and a vendor; rates and defensibility are negotiated better before emergency.
4. Train custodians annually on holds and the consequences of deletion — individual exposure includes criminal obstruction in extreme cases.
5. When litigation arrives: issue the hold immediately, suspend all auto-deletion, document every step, and audit compliance during the case.

Is e-discovery cost proportionate for smaller companies?

It has become survivable. Cloud review platforms priced by matter rather than enterprise license, TAR reducing review volumes by order of magnitude, and the proportionality limits of the 2015 amendments — which curbed the bottomless requests of the 2000s — bring mid-size disputes into manageable ranges. What has not cheapened is the failure mode: the smallest company that deletes covered evidence after a hold faces the same adverse-inference doctrine as the largest. In e-discovery, as in most of litigation, the expensive part is never the process; it is the mistake.